基于形态学特征的机械零件焊接裂纹无损图像检测方法  

Nondestructive Image Detection Method for Welding Cracks of Mechanical Parts Considering Morphological Characteristics

作　　者：刘石桥 LIU Shiqiao(China Nuclear Power Operation Technology Corporation,Wuhan 430205,China)

机构地区：[1]中核武汉核电运行技术股份有限公司,湖北武汉430205

出　　处：《电焊机》2024年第7期31-37,92,共8页Electric Welding Machine

基　　金：中核武汉核电运行技术股份有限公司“智能维修关键技术研发”(B190745)。

摘　　要：在机械零件焊接裂纹检测中,由于采集的图像中焊接边界受到反射波干扰,其存在模糊化噪点,导致检测不准确。为解决该问题,基于边界形态中形状基不变原理,提出考虑形态学特征的机械零件焊接裂纹无损图像检测方法。该方法采用机器视觉技术采集机械零件焊接的视觉图像,通过照明反射模型和傅里叶变换,滤除图像受到光学干扰产生的模糊化噪点。通过二层小波分解图像,构建低频近似分量和高频细节分量。引入形态学算法,以结构元素为核心,通过腐蚀和膨胀处理,并且通过优化腐蚀次数和同态滤波进行优化,提取机械零件焊接裂纹的形态学特征,实现机械零件焊接裂纹无损检测。经过试验证明,所提方法采集的视觉图像更加清晰,可以有效提取裂纹形态学特征和裂纹,其机械零件焊接裂纹检测位置与实际位置基本一致,误差仅为0.01,说明该方法检测精度较高。In the welding crack detection of mechanical parts,because the welding boundary is interfered by reflection wave in the acquired image,there is fuzzy noise,which leads to inaccurate detection.In order to solve this problem,based on the principle of shape basis invariance in boundary morphology,a non-destructive image detection method for welding cracks of mechanical parts considering morphological characteristics was proposed.In this method,machine vision technology is used to collect the visual image of mechanical parts welding,and the fuzzy noise caused by optical interference is filtered by illumination reflection model and Fourier transform.The low frequency approximate component and high frequency detail component are constructed by the two-layer wavelet decomposition.By introducing morphological algorithm,taking structural elements as the core,through corrosion and expansion treatment,and optimization by optimizing corrosion frequency and homomorphic filtering,the morphological characteristics of welding cracks of mechanical parts are extracted,and the non-destructive detection of welding cracks of mechanical parts is realized.The experimental results show that the visual image collected by the proposed method is clearer,and the crack morphological characteristics and cracks can be effectively extracted.The detection location of welding cracks in mechanical parts is basically consistent with the actual location,and the error is only 0.01,indicating that the detection accuracy of the proposed method is high.

关 键 词：机械零件焊接 机器视觉 同态滤波 图像校正 小波变换 形态学特征 

分 类 号：TG441.7[金属学及工艺—焊接]